Calibration of a Terrestrial Full Waveform Laser Scanner Baltimore, Maryland March 27, 2013 Preston J. Hartzell (pjhartzell@uh.edu) Craig L. Glennie Department of Civil and Environmental Engineering University of Houston, Houston, TX 77204 David C. Finnegan Cold Regions Research and Engineering Lab, Hanover, NH 03755
Agenda 1. Introduction & Motivation 2. Method 3. Dataset 4. Results & Analysis System Response Model Absolute Reflectance Incidence Angle Waveform Fitting 5. Future Direction 2
Introduction: Full Waveform LiDAR LiDAR = Light Detection And Ranging Discrete LiDAR XYZ, Intensity Full Waveform LiDAR Digitized record of return echo waveform Derive XYZ, Peak Amplitude (Intensity) Full Waveform Discrete Return Laser Pulse 3
Introduction: Full Waveform LiDAR Primarily used to extract additional echoes (e.g., Gaussian fitting) Gaussian Model Gaussian Models fit to Waveform Fitting parameters also used for target property extraction Most research has relied upon airborne sensors Terrestrial full waveform laser scanners allow control of illuminated targets, enabling more more robust calibration methods 4
Motivation for Calibration Potential: Each digitized echo waveform contains information about the target radiometric and geometric properties Waveform Intensity/Amplitude Discrete intensity and digitized waveform amplitude scales are arbitrary No link between these scales and a target s physical reflectance value Radiometric Calibration: Establishes a link between peak waveform amplitude and target reflectance Reflectance value at laser wavelength provides target spectral property information 5
Motivation for Calibration Potential: Each digitized echo waveform contains information about the target radiometric and geometric properties Waveform Fitting Parametric model fitting struggles with complex system response shapes Sub-optimal models produce improper parameter extraction and decreased ranging accuracy (examples on next slide) System Response Calibration: Determines the system response shape for ideal waveform fitting Analysis of deviations from fitted system response models may yield information on target geometric properties 6
Motivation for Calibration Optech Gemini Riegl VZ-400 Gaussian Model Fit ~ System Response Fit 7
Method Radiometric Calibration Goal = link peak waveform amplitude to target reflectance 1. Store peak waveform amplitude from targets of known reflectance 2. Measurements at multiple ranges Range attenuation System nonlinearity ~Atmospheric attenuation 3. Result = Table of Amplitude vs. Reflectance vs. Range Empirical, system specific Estimate target reflectance from any subsequent waveform 8
Method System Response Calibration Goal = System response model spanning dynamic range 1. Utilize the radiometric calibration measurements 2. Store average waveform Same target, normal incidence Span detector dynamic range 3. Result = Table of system response waveform shapes Empirical, System Specific Accommodates system non-linearity Ideal fitting model for any amplitude waveform return Note Ideally, the emitted laser pulse is sampled Practical alternative is to measure target with known, standard properties 9
Dataset Riegl VZ-400 with full waveform option 1550 nm laser wavelength ~ 5-7 ns FWHM pulse width Two modes: Long Range & High Speed Two detector channels: High power & Low power Labsphere Spectralon: 20% (32%), 50% (60%), 99% reflectivity 10
Dataset Measurement ranges: 2-32 m: 2 m increments 35-70 m: 5 m increments 80-140 m: 10 m increments 160-260 m: 20 m increments Incidence angles 0, 20, 40, 60 99% Spectralon only Sparse waveforms 500 MHz (2 ns) digitizer 5-7 ns pulse width Typical VZ-400 Waveform 11
Results: System Response Model 1. 2. 1. Sparse single waveform 2. Several thousand aligned waveforms at each range 3. Each aligned set of waveforms averaged and combined to form a system response model 3. 12
Results: System Response Model Nonlinear response at high amplitudes Stored in table format for simple interpolation at any desired amplitude or range 13
Results: Absolute Reflectance Peak waveform amplitudes are not proportional to absolute reflectance until beyond ~175 m Estimated target reflectance for a given range and peak amplitude is easily interpolated 14
Results: Incidence Angle 1. 2. 1. Peak amplitude does not follow cosine law until range > ~150 m 2. Increased incidence angle does not increase waveform width Small beam divergence/footprint Width increase below digitizer resolution Target incidence angle information limited to peak amplitude 15
Results: Least Squares Waveform Fit Problem: The model is a table of numbers (no parameters) Requires assumptions about underlying function model (f) Assumed function model is clearly non-linear Partial derivatives of f w.r.t. amplitude (A) and time location (μ): A f μ = NNNNNNNNNN TTTTTTTT WWWWWWWW = SSSSS TTTTTTTT WWWWWWWW Partial derivatives can be evaluated numerically from the calibrated system response data 16
Results: Least Squares Waveform Fit Each iteration utilizes a unique interpolated system response waveform from the calibrated model Elevated tail accommodated very well by calibrated system response model Enables analysis of deviations from standard system response 17
Future Direction Collect calibration data with additional laser wavelength 1064 nm collection scheduled for May 2013 Examine potential for target identification via spectral analysis System response deviation analysis Echo waveform deviation from fitted system response Analysis for correlation to target properties Multi-target waveforms Accommodate target cross-section for more relevant reflectance values in complex multi-target waveforms 18
Wrap-Up 1. Instrument calibration achieved via stored measurements of standard reflectance targets at multiple ranges. 2. Radiometric calibration enables estimation of target absolute reflectance. 3. System response calibration enables ideal waveform echo fitting for precise ranging and deviation measure. 4. The combination of target reflectance and echo waveform deviation will be investigated for enhanced target identification. 19
Thank you Preston J. Hartzell - pjhartzell@uh.edu Craig L. Glennie - clglennie@uh.edu Department of Civil and Environmental Engineering University of Houston, Houston, TX 77204 David C. Finnegan Cold Regions Research and Engineering Lab, Hanover, NH 03755 Acknowledgment: This research was supported in part by an appointment to the Student Research Participation Program at the U.S. Army Cold Regions Research and Engineering Laboratory (CRREL) administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and CRREL. 20